Abstract
The objective of the present study was to develop and assess the validity of a short form of the Child–Adolescent Perfectionism Scale (CAPS). Two Portuguese samples composed of 756 adolescents were used to cross-validate the factorial structure of a nine-item Child–Adolescent Perfectionism Scale—Short Form (CAPS–SF). The CAPS–SF consists of a four-item self-oriented perfectionism subscale and a five-item socially prescribed perfectionism subscale. Both subscales demonstrated adequate internal consistency. Confirmatory factor analysis (CFA) of the CAPS–SF supported the same two-factor structure and represented a very good fit to the data for both groups. Other analyses found that the brief measure of socially prescribed perfectionism was associated with measures of distress and forms of self-criticism. The CAPS–SF appears to represent a reliable and valid alternative to the original CAPS. Overall, the CAPS-SF is considerably briefer than the original CAPS and it offers an economical and valid alternative when measuring perfectionism in children and adolescents.
Perfectionism can be conceptualized as a personality trait leading individuals to develop certain cognitive processes, action tendencies, and behaviors such as the setting of unrealistically high standards for performance accompanied by tendencies for overly critical self-evaluations of one’s behavior (Frost, Marten, Lahart, & Rosenblate, 1990; Stoeber & Janssen, 2011). It was conceptualized initially as a unidimensional construct, limited to intrapersonal aspects. In contrast, Hewitt and Flett (1991) developed a multidimensional conceptualization of perfectionism, assessing three dimensions: self-oriented, socially prescribed, and other-oriented perfectionism. It has been suggested that some dimensions of perfectionism may have some adaptive value (Stoeber et al., 2006), whereas other dimensions have shown close relations with shame (Noble, Gnilka, Ashby, & McLaulin, 2017), self-criticism (Chen, Hewitt, & Flett, 2015), psychological maladjustment, and mental health problems such as anxiety, depression, and disordered eating in adults and adolescents (Bento, Pereira, Roque, Saraiva, & Macedo E Santos, 2017; Blankstein, Lumley, & Crawford, 2007; Hill, McIntire, & Bacharach, 1997; Macedo et al., 2007; Minarik & Ahrens, 1996).
The Child–Adolescent Perfectionism Scale (CAPS) is the most widely used measure when assessing perfectionism in children and adolescents from a multidimensional perspective (Flett et al., 2016). The CAPS consists of 22 items with 12 items assessing self-oriented perfectionism and 10 items assessing socially prescribed perfectionism (Flett et al., 2016). Extensive evidence attesting to its positive psychometric properties and utility has been provided (for a summary, see Flett et al., 2016). The Portuguese version of the CAPS presents adequate psychometric properties, similar to the original version (Bento, Pereira, Saraiva, & Macedo, 2014).
There are several reasons to develop a brief version of the CAPS. In most instances, the use of the original version is preferable but there are times when a briefer and more economical version of the instrument would be beneficial, especially when it is included with several other measures or when there are participants who may have some limitations in their reading ability. The development of a brief version of the CAPS would be particularly useful if the abbreviated version has comparable psychometric characteristics when completed by younger children versus older children and adolescents.
Our decision to create a brief version of the CAPS is also guided by the recognition that some previous researchers have already taken it upon themselves to use a version of the CAPS with fewer than 22 items due to the particular needs of their study. Shorter versions of the CAPS have been utilized with some researchers dropping only a few items (McVey, Pepler, Davis, Flett, & Abdolell, 2002), whereas other researchers have used much fewer items. For instance, Kenney-Benson and Pomerantz (2005) conducted research on maternal control and children’s perfectionism. They focused on a sample of children who were 7 to 10 years old. Perfectionism with the CAPS was assessed with two eight-item subscales measuring self-oriented and socially prescribed perfectionism. The two subscales had high levels of internal consistency but in this sample, there was a much higher correlation between the two perfectionism dimensions (r = .76) than is typically the case in samples that have used the entire CAPS. Another investigation by Saling, Ricciardelli, and McCabe (2005) examined perfectionism in 326 Australian children in Grades 3, 4, and 5. They used an eight-item version of the CAPS that was based on a random selection of four items from each subscale. Unfortunately, the internal consistency was relatively low for self-oriented perfectionism (α coefficients of .51 for boys and .61 for girls), so the authors opted to sacrifice the multidimensional focus and combine all eight items into one overall perfectionism score. This resulted in internal consistencies of .81 for both boys and girls. More recently, a team of investigators examined the ability of the CAPS to assess differences in perfectionism in the sports context (Donachie, Hill, & Hall, 2018). Psychometric analyses resulted in a 10-item version with four items assessing self-oriented perfectionism and six items assessing socially prescribed perfectionism.
Other authors have focused on shorter versions of the CAPS as a result of their efforts to explore the CAPS factor structure. McCreary, Joiner, Schmidt, and Ialongo (2004) analyzed the item responses of Black children from the Baltimore area. They described a 14-item version with a socially prescribed perfectionism subscale and two self-oriented perfectionism factors. As noted by Flett et al. (2016), these results can be questioned because McCreary et al. (2004) used a nonstandard, unique version of the CAPS by significantly altering the measure. Specifically, five CAPS items were reworded. Also, McCreary et al. (2004) decided to use a four-option response key that no longer resembled the standard 5-point response key. Subsequently, O’Connor, Dixon, and Rasmussen (2009) sought to determine whether the three-factor solution could be replicated in their samples of 624 and 737 Scottish adolescents. O’Connor et al. deemed that the three-factor solution described by McCreary et al. (2004) was a poor fit to their data. Instead, they developed an alternative 14-item version. The negatively worded CAPS items were dropped as well as other items that loaded complexly with cross-loadings on the factors. The factors were as follows: (a) a three-item self-oriented perfectionism striving subscale, (b) a four-item self-oriented perfectionism self-critical subscale, and (c) a seven-item socially prescribed perfectionism subscale. These factors were found in both of their samples. Flett et al. (2016) observed that this version is problematic for two reasons. First, at a conceptual level, self-oriented perfectionism is best considered a unidimensional entity, so the theoretical basis for two self-oriented perfectionism factors needs to be explicated. Second, Flett et al. (2016) suggested that the four-item self-critical perfectionism factor identified by O’Connor et al. (2009) is subject to other interpretations due to item content. The four items in this factor seem to generally reflect reactions and responses to imperfection and failure—only one item has wording that openly reflects self-criticism.
Recent developments in the assessment of perfectionism continue to attest to the possibility and suitability of developing short forms of lengthy measures (e.g., Burgess, Frost, & DiBartolo, 2016). Accordingly, we set out to develop a shorter version of the CAPS with one factor representing self-oriented perfectionism and another factor representing socially prescribed perfectionism. As noted below, items were selected based on their association with the full set of items but also in terms of the degree to which the item content reflects the essence of descriptions of self-oriented perfectionism and socially prescribed perfectionism.
The current research primarily explores the feasibility of creating a brief version of the CAPS in a sample of children and adolescents in Portugal. Such a short version might be particularly useful in settings where time constraints make the use of the long form less feasible or advisable (e.g., time-consuming survey research, therapy process-outcome research, and individual monitoring in daily clinical practice).
The research described below outlines the development of the brief CAPS. We also included measures that should be associated with elevated levels of perfectionism in children and adolescents. For instance, given the established relationships between perfectionism and depression, anxiety, and stress (e.g., Hewitt et al., 2002; Hewitt, Flett, & Ediger, 1996), we used the Depression, Anxiety, and Stress Scale–21 Items (DASS-21) to explore the utility and validity of the brief CAPS.
Method
Sample
Participants consisted of 756 adolescents from public and private of middle and high schools of Coimbra, whose parents accepted their adolescents to participate. The mean age was 13.22 years ± 2.28 (range: 9-18). The total sample was divided in two samples considering the elementary school (youngest) and secondary school (oldest).
Sample 1 (S1) consisted of 427 adolescents (241 girls [56.4%] and 186 boys [43.6%]); and Sample 2 (S2) consisted of 329 participants (164 girls [50.1%], 165 boys [49.9%]). All participants took part without financial compensation or any other benefit. The participants in S1 were significantly younger than the participants in S2, S1: 12.31 ± 2.08 versus S2: 13.95 ± 2.10; t(707.498) = 10.676, p < .001.
Materials and Procedure
Procedure
Permission was obtained from the Ethic Commission of Faculty of Medicine of Coimbra, from the Portuguese Data Protection Authority and from the schools headmasters. The informed consent was also obtained from the parents of the adolescents who took part. Confidentiality was ensured.
Measures
The CAPS is a 22-item scale. Responders rated each statement on a 5-point Likert-type scale ranging from 1 (false- not at all true for me) to 5 (very true for me). CAPS scores can range from 22 to 110. Items 3, 9, and 18 from the CAPS are reversed.
Development of the child–adolescent perfectionism scale–short form (CAPS-SF)
The initial version of the short form of the CAPS-SF was constructed with a Portuguese-speaking sample. It was created based on the Portuguese original CAPS. Following Stöber and Joormann’s (2001) procedure for creating short forms of longer self-report measures, we selected the items from the Portuguese original CAPS that demonstrated loadings higher than .65 (Bento et al., 2014). As noted by Stöber and Joormann (2001), these requirements helped to ensure that the short form would evidence a high correlation with the long form of the scale and the items would be representative of their intended subscale domains. We also retained items that were deemed by the authors to capture the essence of self-oriented and socially prescribed perfectionism as psychological constructs. For instance, the items tapping self-oriented perfectionism reflected not only exceptionally high standards, but also generalization tendencies of young people with this orientation (e.g., wanting to be the best at everything), and the sense of personal imperative of them being perfect. Overall, the brief version had four self-oriented perfectionism (SOP) items and five socially prescribed perfectionism (SPP) items. Note that all items from the CAPS-SF subscales were not reverse-coded. This assists children with reading difficulties.
The items that were selected to form the CAPS-SF are listed in Table 1.
Items for the CAPS-SF, Including Item Correlations With Subscale Score (Both the Long and Short Versions).
Note. Values for Sample 2 are shown in parentheses. CAPS-SF = child–adolescent perfectionism Scale–short form; CAPS-LF = child–adolescent perfectionism scale–long form (Bento, Pereira, Saraiva, & Macedo, 2014); α = Cronbach’s alpha.
Sample 1
Sample 2.
DASS-21
The scale comprises 21 items designed to measure the extent to which participants have experienced depressive (D), anxiety (A), and stress (S) symptoms during the past week (Lovibond & Lovibond, 1995). The Portuguese version was used (Ribeiro, Honrado, & Leal, 2004).
The forms of self-criticizing and reassuring scale (FSCRS)
This 22-item self-report scale measure forms of self-criticism and self-reassuring (Gilbert, Clarke, Hempel, Miles, & Irons, 2004). The Portuguese version (Castilho & Gouveia, 2011) was used.
The other as shamer for adolescents–short form (OAS-SF)
The Portuguese version was used (Cunha, Xavier, Cherpe, & Gouveia, 2017; Goss, Gilbert, & Allan, 1994). This self-report measure consists of eight items designed to measure external shame (i.e., what the subjects think about the way others see them).
Psychometric data evaluation
Given the broad age range of the complete sample, we decided to conduct a multigroup analysis, dividing the sample into two age categories. This analysis was conducted to evaluate the model for multigroup invariance of the CAPS-SF across two youngsters groups. The invariance of the structural model for both groups was tested through the chi-square difference test and the critical ratios for differences among all parameter estimates. Significant differences between groups exist when critical ratio values are larger than 1.96 (Byrne, 2013).
To assess overall model fit, a number of goodness of fit measures and recommended cut points were used (Kline, 2005): Chi-Square (χ2), Normed Chi-Square (χ2/df), Comparative Fit Index (CFI ≥ .90, acceptable, and ≥.95, desirable; (Hu & Bentler, 1998), Tucker–Lewis Index (TLI ≥ .90, acceptable, and ≥ .95, desirable; Hu & Bentler, 1998), Goodness of Fit Index (GFI ≥ .90, good, and ≥.95, desirable; Jöreskog & Sörbom, 1996), Root Mean Square Error of Approximation (RMSEA ≤ .07, good fit if GFI ≥ .92; Hair, Black, Babin, & Anderson, 2009) with a 95% confidence interval (CI).
Temporal stability was analyzed by the test–retest correlation method (Pearson correlation). A subset of 341 adolescents (213 girls; mean age = 14.17 years and 132 boys; mean age = 13.46 years) from the original sample completed the CAPS-SF a second time 6 weeks later.
Results
Exploratory Data Analysis
Mean and standard deviations
As the original CAPS, the CAPS-SF rates each statement on a 5-point Likert-type scale ranging from 1 (false—not at all true for me) to 5 (very true for me).
For the total sample, the mean scores were as follows: SOP–short form (SOP-SF): M = 14.41, SD = 3.575; SPP–short form (SPP-SF): M = 14.31, SD = 5.258.
No significant mean differences were observed between both samples for SOP-SF: S1: 14.27 ± 3.734 versus S2: 14.75 ± 3.169; t(–1.803) = 509.430, p = .072 or for SPP-SF: S1: 14.41 ± 5.346 versus 14.13 ± 5.025; t(.694) = 467.269, p = .488.
Internal consistency
Table 1 presents internal consistency reliabilities (using Cronbach’s α) for the CAPS–long form (CAPS-LF; Bento et al., 2014) and CAPS–SF, including the total score and the subscale scores. Internal consistency of the CAPS–SF was higher (S1 α = .84; S2 α = .86) than the CAPS-LF (α = .81). It was found in the total sample that the correlations between the CAPS-SF subscales were r = .77 for self-oriented and r = .90 for socially prescribed; and the correlation between self-oriented and for socially prescribed was r = .43 (p < .01). For the youngest sample, the correlations were as follows: r = .79 for self-oriented and r = .91 for socially prescribed; the correlation between self-oriented and for socially prescribed was r = .46 (p < .01). The correlations in the oldest sample were as follows: r = .72 for self-oriented, r = .90 for socially prescribed and the correlations between self-oriented and for socially prescribed were r = .33 (p < .01).
Both samples (S1 and S2) showed the same factor structure (Table 1).
Test–retest reliability
For the CAPS-SF, the test–retest correlation coefficient was .67 (p < .001) for the total score. The test–retest correlation for the socially prescribed perfectionism subscale was .61 (p < .001) and for self-oriented perfectionism was .67 (p < .001).
Confirmatory Data Analysis
Preliminary data analyses
Analyses for normal distribution did not show any severe bias, as all the variables presented acceptable values of skewness and kurtosis (SK < |3| and Ku < |8-10|) and variance inflation factor (VIF) < 5 (Kline, 2005), which excludes the existence of multicollinearity. Also, from the analysis of multivariate outliers using Mahalanobis distance statistic (D2), some cases indicated the presence of outliers, and were therefore excluded.
Multiple-group confirmatory factor analysis (CFA)
A multigroup analysis was conducted to verify if there were differences in the final model between two groups of younger versus older participants (younger ones from 9 to 15 years old; older ones from 16 to 18). The tested model presented a very good fit to the data for both groups: χ2(24) = 89.420, p < .001; CFI = .980; TLI = .970; RMSEA = .061, CI = [.048, .074]; p = .087. Measurement invariance is suggested when measurement properties are structurally equivalent in different groups (Byrne, 2013). The multiple-group CFA invariance was verified by comparing the unconstrained model (i.e., with free structural parameter coefficients) and the constrained model (i.e., where the parameters are constrained equally across groups; Byrne [2013]). The model presented a very good fit to the data for both groups: GFI = .96; CFI = .975; TLI = .96; RMSEA = .047, p (RMSEA ≤ .05) = .977, 90% CI = [0.037, 0.057].
The multigroup analysis confirmed the invariance of measurement across groups. The unconstrained model (i.e., with free structural parameter coefficients) and the constrained model (i.e., where the parameters are constrained equal across groups) were compared (Byrne, 2013) and results from the chi-square difference test revealed that the model was invariant for both groups for measurement weights (i.e., equal factor loadings), presenting a nonsignificant probability value,
Standardized loadings and correlations for the two-factor model according to the modification indices and theoretical considerations.
Divergent Validity
Scores on the abbreviated socially prescribed subscale presented modest but positive correlations with anxiety (r = .212; p < .01), depression (r = .234; p < .01), and stress (r = .223; p < .01). Scores on this subscale also had significant associations with Inadequate Self (r = .262; p < .01), Hated Self (r = .183; p < .01), and OAS-SF (r = .243; p < .01). There was also a small but significant negative correlation with reassured self (r = –.183; p < .01). In contrast, the abbreviated self-oriented subscale had no significant associations with DASS-21 dimensions, but there was a small but significant weak correlation with reassured self (r = –.103; p = .04) (Table 2).
Correlations Between CAPS-SF, FSCRS Dimensions, and OAS-SF.
Note. CAPS-SF = child–adolescent perfectionism scale–short form; SPP-SF = socially prescribed perfectionism–short form; SOPSF = self-oriented perfectionism–short form; FSCRS-SI = forms of self-criticizing and reassuring scale–self inadequate; FSCRS-SR = forms of self-criticizing and reassuring scale–self-reassured; FSCRS-SH = forms of self-criticizing and reassuring scale–self-hated; OAS-A-SF = other as shamer for adolescents–short form.
p < .05. **p < .01.
Discussion
Over the last three decades, there has been a worldwide growing interest in perfectionism from childhood to adulthood. Because perfectionism has been associated with a wide range of psychopathological conditions, particularly in young ages, the study of this construct can have valuable implications in prevention (see Affrunti & Woodruff-Borden, 2014; Flett & Hewitt, 2014; Teixeira, Pereira, Marques, Saraiva, & Macedo, 2016). There is substantial interest, especially when assessing children and youth, in constructing short form scales that enable researchers to assess a large number of constructs in a reasonably short test session, enhancing the adhesion of participants. In addition, the use of short forms can ensure that measures of many potentially critical variables are obtained so the researcher has greater latitude in testing alternative hypotheses (Widaman, Little, Preacher, & Sawalani, 2011). Accordingly, the present study sought to develop a short version of the CAPS as an extension of previous work on the Portuguese CAPS (Bento et al., 2014) with the expectation that the measure developed could be a short form that is used globally in research in general with participants from North America and other locations.
Our findings can be summarized as follows. First, the short form shows adequate psychometric characteristics with an internal consistency values for the CAPS–SF subscales which were comparable with the internal consistencies for the subscales of the longer CAPS. As Nunnally and Bernstein (1994) mentioned, short form scales reliabilities of .80 or above are generally quite acceptable values.
Second, although the number of scale items was reduced substantially (i.e., nine items instead of 22 items in total), the shortened version had no substantial loss in terms of internal consistency for total scores and after factor analysis; in fact the CAPS-SF yielded the anticipated factor structure. The four items and five items for the subscales have clear content and face validity, but it is also the case that the these items had, almost without exception, the highest factor loadings in the factor analysis of the CAPS items reported in Flett et al. (2016).
Third, CFA confirmed the proposed structure with good fit indices also showing that the shortened CAPS have the same higher order factor structure as the original full scale. Fourth, the multigroup analysis confirmed the invariance of measurement across age groups, which supports that this scale is appropriate to use across the entire age range.
Fifth, the temporal stability reliability (SOP-SF: r = .67 and SPP: r = .61) was comparable with the Portuguese original version (SOP: r = .69 and SPP: r = .59). The test–retest reliabilities were quite comparable with the 1-year test–retest reliabilities for the two original CAPS subscales (.65 for self-oriented perfectionism, .59 for socially prescribed perfectionism) when the full CAPS was administered to a large sample of adolescents from Canada.
Finally, the CAPS-SF subscales had modest associations with the DASS-21 dimensions. Specifically, the abbreviated measure of socially prescribed perfectionism was associated significantly with all three DASS-21 indices, but there was little association between these measures and self-oriented perfectionism. This pattern of findings is not in keeping with recent research by Sironic and Reeve (2015) who found that both self-oriented and socially prescribed perfectionism were associated positively and significantly with the DASS-21 subscale measures of anxiety, depression, and stress in their sample of Australian adolescents.
The results with the measures of self-criticism and being shamed by others showed similarly that the brief measure of socially prescribed perfectionism was linked with a self-critical orientation and reports of being shamed by others, and these associations were not found with the abbreviated measure of self-oriented perfectionism. We interpreted these results as illustrating the associated constructs that contribute to the vulnerability inherent in socially prescribed perfectionism; however, the modest magnitude of the obtained associations further illustrates that the perfectionism construct as assessed by the abbreviated CAPS does not overlap to too much of an extent with related constructs such as self-criticism and shame.
Clearly, the current findings point to several directions for future research. First and foremost, the factor structure and other psychometric features of the abbreviated CAPS subscale need to be examined in other samples. Given that the nine-items in the abbreviated CAPS were also items that were also found as having the strongest loadings in the analyses conducted by Flett et al. (2016), we are confident that generalizability will be supported to a substantial degree. It is also evident that research is needed to further explore the nomological network associated with the brief assessment of self-oriented perfectionism and socially prescribed perfectionism in children and adolescents.
Collectively, the present findings attest to the potential usefulness of the abbreviated nine-item CAPS. Our analyses established that there is clear evidence of the multidimensionality of trait perfectionism in children and adolescents with this version of the CAPS. The obtained evidence from our Portuguese speaker samples suggested that this version can be used both effectively and efficiently as an economical alternative to the full CAPS. The CAPS-SF addresses the need for a brief multidimensional inventory for assessing trait perfectionism in children and adolescents, and its characteristics support the contention that this version of the measure merits consideration by other researchers as an alternative to the practice of researchers selecting their own subset of items.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Portuguese General Health Direction included in the Project “Emotional and Behavioral deregulation in a School Population Sample.”
